Predictive Carbon Footprint Analytics in Sustainable Logistics: Assessing Explainable AI (XAI) Frameworks for Proactive Emission Penalty Mitigation across Multi-Tiered Supply Networks

• Billy Elly

  Lautech

  Author

The global intensification of climate regulations, exemplified by the International Maritime Organization's Net Zero Framework imposing penalties of $380 per ton of CO₂ deficit and the European Union's Carbon Border Adjustment Mechanism, has created urgent compliance imperatives for logistics networks. However, existing carbon footprint monitoring systems remain constrained by retrospective Life Cycle Assessments and static reporting that fail to enable proactive emission penalty avoidance across multi-tiered supply chains. This study addresses this gap by developing and comparatively evaluating an Explainable AI (XAI) framework that integrates Long Short-Term Memory networks for predictive emission forecasting with SHapley Additive exPlanations for model interpretability, benchmarked against Random Forest and XGBoost classifiers. Analysis of integrated shipment records and country-level energy intensity metrics demonstrated that the XGBoost model achieved the strongest predictive performance with 89.4% accuracy and a recall of 0.76 for high-emission shipments, while SHAP analysis identified shipment size and transport duration as the most influential predictors. The framework's XAI component successfully translated black-box predictions into actionable managerial insights, enabling proactive route optimization and emission penalty avoidance. The study contributes a replicable, scalable decision-support architecture that bridges predictive machine learning, explainability, and sustainability compliance, offering logistics managers a practical tool for embedding decarbonization into daily operational planning.

• PDF

Copyright (c) 2026 Billy Elly (Author)

This work is licensed under a Creative Commons Attribution 4.0 International License.